GLOBAL PERSPECTIVE · INDEPENDENT THINKING
UNDERSTAND WHAT MOVES THE WORLD.
Human Behaviour · Analysis

What You Expect to See Is Not Always What You Need to Judge

A visual-perception experiment separates prediction from task relevance. That distinction suggests a sharper question for training: what exactly should the learner inspect?

BY GABRIEL NICOLAEV – CYGURU · · 5 MIN READ
READING TOOLS
Close-up of fingers pressing keys on a computer keyboard.
THE READING GUIDE

Inside this article

01Recognition is not the whole assignment ↓02Write the criterion before choosing the practice material ↓03Keep the surface familiar while changing what matters ↓

Recognition is not the whole assignment

Imagine an apprentice proofreader studying a page of short product descriptions. The lines share a familiar shape: name, dimensions, material and colour. After several examples, the next line looks ordinary. But the assignment is not simply to recognise an ordinary-looking line. It is to decide whether a particular specification matches the source supplied by the manufacturer. The familiar pattern and the required judgment are different things. A training exercise should make that distinction explicit.

A NeuroImage study by Thomas Holstein and colleagues offers a narrow scientific prompt. Participants judged sequential visual gratings while EEG was recorded; responding depended on a task-relevant colour. The abstract reports faster responses and later processing differences for predicted, relevant stimuli, without the reported early visual modulation. CYPAPER reviewed the indexed primary abstract, not the full paper. The study appeared online on 17 October; 15 November is its issue date. It did not study proofreading, and the training proposals below are editorial applications, not demonstrated effects of its experiment.

Write the criterion before choosing the practice material

A trainer could ask the apprentice to find misspellings, compare dimensions, identify unsupported claims or assess whether an instruction is understandable. Each is a legitimate task. They require different judgments even when the visible paragraph remains unchanged. A single instruction to check the text carefully leaves the learner to infer which kind of error matters most and how much evidence a satisfactory answer requires.

For the hypothetical specification exercise, the criterion could be stated precisely: compare every numerical dimension with the corresponding field in the source sheet, including its unit. A second pass could examine whether the description names the correct material. A third could address wording. Separating these passes would not establish that separation always improves performance. It would make the exercise interpretable: the trainer could see which judgment had been attempted and where a correction was needed.

The same principle applies when selecting examples. If every error is a conspicuous spelling mistake, the exercise cannot tell the trainer much about a learner's handling of an accurately spelled but incorrect material. The examples should include the distinctions the assignment actually asks the learner to make, rather than rewarding success at an easier substitute.

Keep the surface familiar while changing what matters

Consider a deliberately constructed pair of practice descriptions. Both use the same layout and fluent language. One correctly describes a cotton cover; the other calls the cover linen although its source sheet says cotton. A learner who points only to grammatical consistency has answered a different question from the one the exercise posed. The trainer can ask the learner to identify the evidence that settles the material question without treating the missed discrepancy as a diagnosis of carelessness.

A second pair could reverse the relationship. One description uses an unusual phrase but accurately represents the source. Another uses polished, familiar wording but changes a measurement. These are hypothetical training examples. Their value would lie in testing whether the chosen criterion survives changes in how comfortable the text feels to read. The unusual phrase should not be counted as a factual error merely because it interrupts the expected style.

The answer key must preserve those distinctions. Mark factual accuracy, clarity and house style separately. A trainer who silently changes the criterion between examples cannot fairly interpret the learner's changing answers. Nor should an apprentice be penalised for following the stated task while failing to guess an unstated editorial preference.

Ask for a reason that refers to the evidence

After a judgment, invite the learner to show what they compared. For a dimension, that might mean pointing to the number and unit in both documents. For a material, it might mean naming the source field. This explanation creates something the trainer can inspect. It also makes a correction specific: the learner may have consulted the wrong row, mistaken a unit or applied a stylistic rule where factual comparison was required.

CYPAPER's discussion of useful feedback concerns the next decision a recipient can make. In this exercise, useful feedback would identify the comparison that needs to be repeated. An instruction to concentrate harder supplies no equivalent information. A worked correction could show why the original answer looked plausible and which detail changes the conclusion, then offer a different example requiring the same distinction.

The learner should also be allowed to say that the evidence is insufficient. Suppose the source sheet does not specify the cover material at all. Guessing the most likely material is not the same task as verifying the description. A defensible answer would identify the missing fact and ask for it. That response deserves a place in the answer key before the exercise begins.

Change the task deliberately, then check the transfer

A later session could give the same page a different assignment: inspect the wording for ambiguity rather than compare specifications. The trainer would announce the change and ask the learner to explain what now counts as relevant evidence. This would test whether the person can adopt a new criterion, rather than merely repeat the sequence that succeeded in the previous session. It would not measure general intelligence or establish a universal capacity for attention.

A further exercise could introduce a different layout while preserving the original verification task. The trainer could then examine whether the learner still finds the relevant source fields without relying on their former positions. Any conclusion should remain tied to the examples and conditions actually tested. Performance on a short training set does not certify reliable work across an entire publication or business.

The larger educational aim is a form of precision: knowing which question one's judgment answers. Familiarity can tell a learner that they have encountered a pattern before. The assignment still has to specify what should be decided about it. Good training makes that question visible, supplies evidence that can settle it, and gives uncertainty a legitimate response when the material does not support an answer.

Gabriel Nicolaev — CyGuru

Sources & further reading

Holstein and colleagues — indexed abstract and publication dates ↗

Illustrative photograph by Alicia Christin Gerald / Unsplash; not an image of the study or its participants.

FOLLOW THE CONNECTION

Another angle on this subject

Continue through related ideas, decisions and consequences.